New preprint questions whether police workforce analytics can demonstrate fair and useful outcomes

Police standards leads, HR, workforce representatives and oversight bodies need to know whether monitoring improves an existing investigative process. Ask which narrowly defined concern justifies data linkage, how benign explanations are considered and whether a less intrusive process could work. A bounded engagement could compare review workflows using synthetic records before any live expansion. The value hypothesis is clearer accountability and reduced avoidable review burden. This preprint does not prove discrimination, validate a competing vendor or establish a local commercial opportunity.

Fit is a controlled analytics pipeline with explicit purpose permissions and traceable human review. Inventory source quality, ownership and data joins before building features. Test legitimate schedule exceptions, duplicate identities, stale HR data and access attempts across welfare and disciplinary roles. A proof of value should compare reviewers' supported decisions against a predefined reference process and measure unnecessary review work. Do not equate unreviewed alerts with false positives. Infrastructure sizing should follow a bounded workload; the source does not establish a cloud or on-premises preference.

Assign a professional-standards owner and a separate assurance reviewer, supported by HR data specialists and staff representatives. Define permitted uses, notify affected staff, rehearse a contested alert and record escalation ownership.

Proposed acceptance

every pilot alert has a documented disposition, every challenged record has a correction route, and access-control tests pass before live use. Establish local workload and fairness thresholds in advance. Dependencies include usable outcome data and meaningful consultation. Risks include expanding scope during a pilot and undermining welfare participation through ambiguous data reuse.